Professional Documents
Culture Documents
Dr. Arthur Benjamin Weglein-A Timely and Necessary Antidote To Indirect Methods and So-Called P-Wave FWI.
Dr. Arthur Benjamin Weglein-A Timely and Necessary Antidote To Indirect Methods and So-Called P-Wave FWI.
Dr. Arthur Benjamin Weglein-A Timely and Necessary Antidote To Indirect Methods and So-Called P-Wave FWI.
October 2013
Downloaded 04/10/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
R o u n d t a b l e
(1)
(2)
(4)
when |S/a|. < 1
If, rather than these nonlinear relationships among S, a,
and r, we instead imagine an exact linear relationship that S,
a, and r might satisfy, e.g.,
,
(5)
(6)
1193
Downloaded 04/10/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
R o u n d t a b l e
October 2013
Downloaded 04/10/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
R o u n d t a b l e
phase so that we can call that model-matching scheme fullwaveform inversion. Why cant we solve for density and velocity uniquely from a single trace, because we can certainly
model the single trace from knowing the velocity and density
as a function of depth? Thats a beginning and an example of
thinking that solving a forward problem in an inverse sense is
in some way actually solving the inverse problem. What came
along in that earlier time, as a response to this question, were
direct acoustic inversion methods that said that inverting
for velocity and density as functions of depth from a single
trace is impossible, or at least that it is impossible to provide the unique and actual velocity and density as a function
of depth. That direct-inversion framework convinced many
(hopefully most) people that the one-trace-in, two-functionsout approach is not a question or an issue of which indirect
algorithm or LP cost function you are using. It is more basic and stands above algorithm; its an inadequate-data issue.
No algorithm with that single-trace data input should call
itself inversion, even if that single trace was model-matched
and iteratively updated and computed with amplitude and
phase and, with too much self-regard, labels itself as fullwave inversion. We learned to stop running that single trace
through search algorithms for velocity and densityand that
lesson was absorbed within our collective psyches in our industryfor whatever the cost function and local or global
minimum you employed. Using the wrong and fundamentally inadequate data closes the book and constitutes the end of
October 2013
1195
Downloaded 04/10/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
R o u n d t a b l e
October 2013
(8)
(10)
(12)
... .
A unique expansion of VG0 in orders of measurement values of (G-G0 ) is
(13)
Downloaded 04/10/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
R o u n d t a b l e
The ability of the forward series to have a direct inverse series derives from (1) the identity among G, G0 , V provided by
the scattering equation and then (2) the recognition that the
forward solution can be viewed as a geometric series for the
data, D, in terms of VG0. The latter derives the direct inverse
series for VG0 in terms of the data.
Viewing the forward problem and series as the Taylor
series (Equation 7) in terms of m does not oer a direct
inverse series, and hence there is no choice but to solve the
forward series in an inverse sense. It is that fact that results
in all current AVO and FWI methods being modeling methods that are solved in an inverse sense. Among references that
solve a forward problem in an inverse sense in P-wave AVO
are Beylkin and Burridge (1990), Boyse and Keller (1986),
October 2013
1197
Downloaded 04/10/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
R o u n d t a b l e
Burridge et al. (1998), Castagna and Smith (1994), Clayton and Stolt (1981), Foster et al. (2010), Goodway (2010),
Goodway et al. (1997), Shuey (1985), Smith and Gidlow
(2000), Stolt (1992), and Stolt and Weglein (1985). The intervention of the explicit relationship among G, G0, and V
(the scattering equation) in a Taylor series-like form produces
a geometric series and a direct inverse solution.
The linear equations are:
(23)
(24)
(14)
(25)
(26)
(15)
(27)
(16)
relates to
,
relates to
, and so on,
Because
the four components of the data will be coupled in the nonlinear elastic inversion. We cannot perform the direct nonlinear
inversion without knowing all components of the data. Thus,
the direct nonlinear solution determines the data needed for
a direct inverse. That, in turn, denes what a linear estimate
means. That is, a linear estimate of a parameter is an estimate
of a parameter that is linear in data that can directly invert for
that parameter. Because DPP , DPS , DSP , and DSS are needed to
determine a , a , and a directly, a linear estimate for any one
of these quantities requires simultaneously solving Equations
1922. See, e.g., Weglein et al. (2009) for further detail.
Those direct nonlinear formulas are like the direct solution
for the quadratic equation mentioned above and solve directly
and nonlinearly for changes in VP , VS, and density in a 1D elastic
Earth. Stolt and Weglein (2012), present the linear equations
for a 3D Earth that generalize Equations 19-22. Those formulas
prescribe precisely what data you need as input, and they dictate how to compute those sought-after mechanical properties,
given the necessary data. There is no search or cost function, and
the unambiguous and unequivocal data needed are full multicomponent dataPP, PS, SP, and SSfor all traces in each
of the P and S shot records. The direct algorithm determines
rst the data needed and then the appropriate algorithms for
using those data to directly compute the sought-after changes in
the Earths mechanical properties. Hence, any method that calls
itself inversion (let alone full-wave inversion) for determining
changes in elastic properties, and in particular the P-wave velocity, VP , and that inputs only P-data, is more o base, misguided,
and lost than the methods that sought two or more functions of
depth from a single trace. You can model-match P-data until the
cows come home, and that takes a lot of computational eort
and people with advanced degrees in math and physics computing Frechet derivatives, and requires sophisticated LP norm
cost functions and local or global search engines, so it must be
reasonable, scientic, and worthwhile. Why cant we use just
PP data to invert for changes in VP , VS , and density, because
Zoeppritz says that we can model PP from those quantities, and
because we have, using PP-data with angle variation, enough
dimension? As stated above, data dimension is good, but not
good enough for a direct inversion of those elastic properties.
(17)
(18)
(19)
(20)
(21)
and
(22)
where a (1), a(1), and a(1) are the linear estimates of the changes
in bulk modulus, shear modulus, and density, respectively.
The direct quadratic nonlinear equations are
1198
October 2013
Downloaded 04/10/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
R o u n d t a b l e
1200
October 2013
Downloaded 04/10/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
R o u n d t a b l e
October 2013
wrong with the approximation? If you purposefully or inadvertently ignore (or wish away) the framework and algorithms
that a direct solution to the elastic parameter estimation provides, you will never know what you are ignoring and dropping
and what your approximation is approximating, nor will you
know what value your method actually represents and means,
and how you could improve the reliability of your prediction.
In summary, so-called P-wave FWI is something less than
advertised and is in general the wrong (acoustic) Earth model,
the wrong data, and the wrong methodbut besides that, it
has a lot going for it.
In Zhang (2006), the direct elastic inverse was applied to
a 4D application and the term beyond linear was able to help
distinguish a pressure change from a uid change. This line
of research continued in Li (2011) and Liang (2010). This is
comparatively illustrated with synthetic log data in Zhangs
Figures 1 through 6 (which are included in this article).
Epilog
A direct method to nd the route from where you are to
where you want to goe.g., for a scheduled meeting
would use MapQuest, while an indirect method would seek
and search and stop at every possible location in the city until
you arrive somewhere where someone seems to be happy to
see you, and you have a toolbox of LP cost functions to dene
happy. A direct solution, in contrast to indirect methods,
does not require or ever raise the issue of necessary but insufcient conditions or cost functions, and its not a condition
or property. Its a solution, a construction. Nothing beats
that for clarity, eciency, and eectiveness. The direct MapQuest inversion communication and message to the current
indirect P-wave FWI methods is that the latter are searching
for the meeting in the wrong city.
The message of this article is that direct inversion provides
a framework, and a set of data requirements and algorithms,
that not only have produced a standalone capability (with
model-type independent algorithms) for removing free-surface and internal multiples, without subsurface information,
but also for establishing the requirements for all seismic processing methods that depend on amplitude analysis, such as
AVO and so-called FWI. Being frank, we wish these requirements were not the case, because it makes our lives more complicated and dicultbut the conclusions are inescapable.
When the framework, data requirements, and direct methods
are not satised, we have a clear and understandable reason
for the resulting failure and for what we might do to provide
more reliable and useful predictive capability. Direct and indirect methods both play an essential role in an eective seismic processing strategy: where the former accommodates the
physics within the system, and the latter provides a channel
for real-world phenomena beyond the assumed physics.
References
Anderson, J. E., L. Tan, and D. Wang, 2012, Time-reversal checkpointing methods for RTM and FWI: Geophysics, 77, no. 4, S93
S103, http://dx.doi.org/10.1190/geo2011-0114.1.
Baumstein, A., J. E. Anderson, D. L. Hinkley, and J. R. Krebs, 2009,
Downloaded 04/10/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
R o u n d t a b l e
Scaling of the objective function gradient for full-waveeld inversion: 79th Annual International Meeting, SEG, Expanded Abstracts, 22432247, http://dx.doi.org/10.1190/1.3255307
Beylkin, G. and R. Burridge, 1990, Linearized inverse scattering problem of acoustics and elasticity: Wave Motion, 12, no. 1, 1522,
http://dx.doi.org/10.1016/0165-2125(90)90017-X.
Ben-Hadj-ali, H., S. Operto, and J. Virieux, 2008, Velocity model
building by 3D frequency-domain, full-waveform inversion of
wide-aperture seismic data: Geophysics, 73, no. 5, VE101VE117,
http://dx.doi.org/10.1190/1.2957948.
Ben-Hadj-ali, H., S. Operto, and J. Vireux, 2009, Ecient 3D frequency-domain full-waveform inversion (FWI) with phase encoding:
71st Conference and Exhibition, EAGE, Extended Abstracts, P004.
Biondi, B. and P. Sava, 1999, Wave-equation migration velocity
analysis: 69th Annual International Meeting, SEG, Expanded Abstracts, 17231726, http://dx.doi.org/10.1190/1.1820867.
Biondi, B. and W. Symes, 2004, Angle-domain common-image
gathers for migration velocity analysis by waveeld-continuation imaging: Geophysics, 69, no. 5, 12831298, http://dx.doi.
org/10.1190/1.1801945.
Blum, E. K., 1972, Numerical analysis and computation: Theory and
practice: Addison-Wesley.
Boyse, W. E. and J. B. Keller, 1986, Inverse elastic scattering in three
dimensions: The Journal of the Acoustical Society of America, 79,
no. 2, 215218, http://dx.doi.org/10.1121/1.393561.
Brandsberg-Dahl, S., M. de Hoop, and B. Ursin, 1999, Velocity
analysis in the common scattering-angle/azimuth domain: 69th
Annual International Meeting, SEG, Expanded Abstracts, 1715
1718, http://dx.doi.org/10.1190/1.1820865.
Brossier, R., S. Operto, and J. Virieux, 2009, Robust elastic frequency-domain full-waveform inversion using the L1 norm:
Geophysical Research Letters, 36, no. 20, L20310, http://dx.doi.
org/10.1029/2009GL039458.
Burridge, R., M. de Hoop, D. Miller, and C. Spencer, 1998, Multiparameter inversion in anisotropic elastic media: Geophysical Journal International, 134, 757777.
Castagna, J. and S. Smith, 1994, Comparison of AVO indicators: A
modeling study: Geophysics, 59, no. 12, 18491855, http://dx.doi.
org/10.1190/1.1443572.
Chavent, G. and C. Jacewitz, 1995, Determination of background
velocities by multiple migration tting: Geophysics, 60, no. 2,
476490, http://dx.doi.org/10.1190/1.1443785.
Clayton, R. W. and R. H. Stolt, 1981, A Born-WKBJ inversion method for acoustic reection data: Geophysics, 46, no. 11, 15591567,
http://dx.doi.org/10.1190/1.1441162.
Crase, E., A. Pica, M. Noble, J. McDonald, and A. Tarantola,
1990, Robust elastic nonlinear waveform inversion: Application to real data: Geophysics, 55, no. 5, 527538, http://dx.doi.
org/10.1190/1.1442864.
Ferreira, A., 2011, Internal multiple removal in oshore Brazil seismic
data using the inverse scattering series: Masters thesis, University
of Houston.
Fitchner, A., 2011, Full seismic waveform modeling and inversion:
Springer-Verlag.
Foster, D., R. Keys, and F. Lane, 2010, Interpretation of AVO
anomalies: Geophysics, 75, no. 5, 75A375A13, http://dx.doi.
org/10.1190/1.3467825.
Gauthier, O., J. Virieux, and A. Tarantola, 1986, Two dimensional
nonlinear inversion of seismic waveforms: Geophysics, 51, no. 7,
13871403, http://dx.doi.org/10.1190/1.1442188.
Goodway, B., 2010, The magic of Lam: The Leading Edge, 29, no. 11,
1432, http://dx.doi.org/10.1190/tle29111432.1.
October 2013
1203
Downloaded 04/10/14 to 129.7.16.11. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
R o u n d t a b l e
October 2013
dx.doi.org/10.1190/1.1442046.
Tarantola, A., 1987, Inverse problem theory: Method for data tting
and model parameter estimation: Elsevier.
Taylor, J. R., 1972, Scattering theory: the quantum theory of nonrelativistic collisions: John Wiley and Sons.
Valenciano, A., B. Biondi, and A. Guitton, 2006, Target-oriented
wave-equation inversion: Geophysics, 71, no.4, A35A38, http://
dx.doi.org/10.1190/1.2213359.
Vigh, D. and E. W. Starr, 2008, 3D prestack plane-wave, full-waveform inversion: Geophysics, 73, no. 5, VE135VE144, http://
dx.doi.org/10.1190/1.2952623.
Weglein, A.B., 2012, Short note: An alternative adaptive subtraction
criteria (to energy minimization) for free surface multiple removal:
M-OSRP 2011 Annual Report, 375.
Weglein, A. B., 2012, Short note: A formalism for (1) modeling the
amplitude and phase of pressure waves from a heterogeneous elastic medium and (2) selecting and predicting P-wave events that
have only experienced pressure-wave episodes in their history: MOSRP 2011 Annual Report, 364370.
Weglein, A. B., S.-Y. Hsu, P. Terenghi, X. Li, and R. H. Stolt,
2011, Multiple attenuation: Recent advances and the road ahead
(2011): The Leading Edge, 30, no. 8, 864875, http://dx.doi.
org/10.1190/1.3626494.
Weglein, A. B., F. Liu, X. Li, P. Terenghi, E. Kragh, J. Mayhan, Z. Wang,
J. Mispel, L. Amundsen, H. Liang, L. Tang, and S.-Y. Hsu, 2012, Inverse scattering series direct depth imaging without the velocity model: rst eld data examples: Journal of Seismic Exploration, 21, 128.
Weglein, A. B., H. Zhang, A. C. Ramirez, F. Liu, and J. Lira, 2009,
Clarifying the underlying and fundamental meaning of the approximate linear inversion of seismic data: Geophysics, 74, no. 6,
WCD1WCD13, http://dx.doi.org/10.1190/1.3256286.
Weglein, A. B. and K. Matson, 1998, Inverse-scattering interval multiple attenuation: an analytic example and subevent interpretation
in S, Hassanzadeh, ed., Mathematical methods in geophysical
imaging: SPIE, 10081017.
Weglein, A. B., F. V. Arajo, P. M. Carvalho, R. H. Stolt, K. H.
Matson, R. T. Coates, D. Corrigan, D. J. Foster, S. A. Shaw,
and H. Zhang, 2003, Inverse scattering series and seismic exploration: Inverse Problems, 19, no. 6, R27R83, http://dx.doi.
org/10.1088/0266-5611/19/6/R01.
Weglein, A. B., D. J. Foster, K. H. Matson, S. A. Shaw, P. M. Carvalho, and D. Corrigan, 2002, Predicting the correct spatial location
of reectors without knowing or determining the precise medium
and wave velocity: initial concept, algorithm and analytic and numerical example: Journal of Seismic Exploration, 10, 367382.
Zhang, H., 2006, Direct nonlinear acoustic and elastic inversion: towards fundamentally new comprehensive and realistic target identication: PhD thesis, University of Houston.
Zhang, Y. and B. Biondi, 2013, Moveout-based wave-equation migration velocity analysis: Geophysics, 78, no. 2, U31U39, http://
dx.doi.org/10.1190/geo2012-0082.1.
Zhou, H., L. Amundsen, and G. Zhang, 2012, Fundamental issues
in full-waveform inversion: 82nd Annual International Meeting, SEG, Expanded Abstracts, http://dx.doi.org/10.1190/segam2012-0878.1.